computer-use

browser-use/browser-use

Python toolkit that lets LLM agents drive real browsers — Browser Use lets AI agents control a browser via Playwright/CDP, with an optional cloud service offering stealth browsers and a hosted agent API. It targets developers building autonomous web agents for scraping, testing, or task automation without managing browser infrastructure themselves.

Mintplex-Labs/anything-llm

A local-first LLM workspace that turns documents into working agents — AnythingLLM bundles RAG chat, multi-user support, and configurable AI agents into one self-hosted app, aimed at people who want to run document-grounded chat and agentic workflows locally instead of relying on hosted AI services. It targets self-hosters and teams needing private, multimodal, agent-capable LLM tooling without complex setup.

bytedance/UI-TARS-desktop

The Open-Source Multimodal AI Agent Stack: Connecting Cutting-Edge AI Models and Agent Infra

trycua/cua

Full desktops for AI agents, with open-source tooling behind it — Cua bundles the pieces needed for computer-use agents: a desktop automation driver, Lume for local VMs on Mac and Windows, CUA-S1 decision models, and Cua Bench for evaluation. It targets developers building or testing agents that need real OS environments rather than simulated ones.

vercel-labs/agent-desktop

Controlling desktop apps through the accessibility tree, not pixels — A CLI and C library that lets AI agents read and operate macOS and Windows apps via OS accessibility trees, outputting JSON with stable refs for each actionable element. It covers 56 operational commands, with risky key/mouse-hold actions blocked until a daemon can safely release them. Linux support is planned.

milind-soni/OpenMausBot

Chat with your own team of local AI agents, Grok Bot style — OpenMausBot is an open-source alternative to Grok Bot, Meta Muse, OpenAI dots and Cue by Manus. Each contact in the sidebar is a real agent powered by Claude or Codex, with its own personality, model, cloud computer and connected apps—self-hosted and bring-your-own-model rather than tied to one vendor's cloud.

NanmiCoder/cc-haha

A desktop command center for Claude Code and AI agent teams — cc-haha wraps Claude Code into a local-first desktop app with multi-agent teams, Git worktrees, Diff review, a skills marketplace, and Computer Use that controls other apps without hijacking your mouse. Remote access via WeChat, Feishu, DingTalk, Telegram, WhatsApp or H5 suits developers managing coding agents across devices.

simular-ai/Agent-S

Agent S: the state-of-the-art open source computer use agents

Anil-matcha/suparagent

A self-hosted, MIT-licensed workspace for Claude Code agent tasks — Open Dots wraps the Claude Code CLI in a self-hosted workspace: run coding tasks from Telegram or a browser, review risky actions before they execute, connect GitHub, and schedule recurring prompts. It's an early prototype, Claude-only, aimed at people evaluating paid agent products like OpenAI Dots or Manus Cue.

onehumanbeing/esp32-iphone-agent

Let an AI model use a real iPhone with a $7 ESP32: AirPlay eyes, USB HID hands, any OpenAI-compatible vision model. No Mac, no developer account, no jailbreak.

arcboxlabs/arcbox

An open-source, Rust-built Docker and VM runtime for Macs — ArcBox is a from-scratch Rust runtime that bundles Docker-compatible containers, native Kubernetes, disposable microVM sandboxes for AI agents, and full Linux/macOS VMs under one daemon. Positioned as an MIT/Apache-2.0 alternative to closed-source Docker Desktop and OrbStack, it runs fully local with no cloud control plane, aimed at Mac developers and agent sandboxing use cases.

nano-muse/nanoMuse

A cross-device personal agent that uses your screen as its hands — nanoMuse is an open-source agent that runs on Android, iOS/iPad, Windows/macOS/Linux and in the browser, controlling the device's GUI to act on your behalf. It keeps working after the app closes, remembers context, and asks permission before irreversible actions, aiming at users who want one consistent agent across all their devices.

daveebbelaar/bama

BAMA: Build Any Mac App. Agent skills to research, clone, design, build, optimize, and test native macOS apps with Swift and SwiftUI.

tembo/voxl

Rust-built remote desktop streaming that beats noVNC on speed — voxl streams X11, macOS or Windows screens as H.264 over WebSocket to a browser client using WebCodecs, with input, clipboard, audio and file transfer built in. Benchmarks show 60 FPS with 12-99ms latency versus under 8 FPS for noVNC, and comparisons against selkies, KasmVNC, Neko and Sunshine. Useful for sandboxed desktops, AI agent control, or software-factory setups needing low-latency browser access.

Amal-David/awesome-jev

A curated map of Jev demos, SDKs, and agent tooling — This awesome-list gathers demos, SDKs, browser/desktop tools, and reading material around Jev, TypeSafe's probability-based yes/no judgment model. It's aimed at developers building typed classifiers or agents, with source links and notes on setup, privacy, and evidence limits rather than just hype.

mrveiss/AutoBot-AI

Self-hosted, agentic AI you own — voice, vision-driven browser & computer control, visual workflows, human-in-the-loop approvals, multi-user RBAC, and a knowledge graph. Runs on your infrastructure. Your data. Your AI.